- Robotic Process Automation (RPA), a type of artificial intelligence, can reduce claim processing times from days to seconds
- By 2030, 90% of personal and small‑business claims will be processed by algorithms
- Chatbots and other AI tools can boost customer satisfaction when used properly
Jim was only four months into his first‑ever assignment as a claims adjuster when he set a world record. After a customer reported his $979 parka from Saks Fifth Avenue had been stolen, Jim cross‑referenced the claim with his policy, ran 18 anti‑fraud algorithms on it and wired a full payout to his bank—all in three seconds.
Jim is no ordinary adjuster. He’s “AI Jim,” a chatbot created by insurance tech startup Lemonade. Jim is just one example of how the $4.6 trillion global insurance industry is overhauling claims processing, one of its most complex business processes.
The shift is seismic. By 2030, 90% of personal and small‑business insurance claims processing will be entirely automated, according to a 2018 McKinsey report. And AI‑powered chatbots will be the primary claims processing touchpoint for most customers, with 70% to 90% fewer human claims personnel than in 2018.
“All of the major traditional insurers recognize the tech revolution that’s happening. They can see the train wreck coming,” says Brian Retzlaff, executive solutions consultant at ServiceNow. “The question is, will they be able to transform in time to stay relevant or will they be devoured by nimbler competitors?”
A fully automated future for insurance is still a decade or so away, according to McKinsey. But Retzlaff’s point is clear: Both major players and VC‑backed upstarts are planting seeds today that will determine who dominates the digital insurance landscape of tomorrow.
Fast track for claims, speed trap for fraud
Since it launched its One Day Pay initiative in 2015, Aflac has paid out more than two million claims within 24 hours using a new online platform. The upside has been dramatic: Customer satisfaction rates rose from 39% to 79% in just three years.
That’s solid progress, but still lags far behind the capabilities of emerging AI and automation tools. For example, China’s Ant Insurance says its new AI platform for claims management can complete a claim in one second. Just two years ago, the process took 49 hours. Insurance giant Prudential recently launched an AI‑powered system in Hong Kong to settle hospital claims, shortening its process from nine days to 2.3 seconds.
These advances are possible because today’s AI applications can quickly identify fraudulent claims. Insurers estimate that fraud costs them $80 billion annually across all types of policies. That’s why more than 75% of them are already using AI tools to fight fraud.
Algorithms scan claims for suspicious inconsistencies or patterns like overlapping networks of doctors or lawyers. Social media analytics uncover connections between known scammers. Insurers are even deploying natural language processing for real‑time “sentiment analysis” of claims calls.
Together these tools help insurers flag cases that require in‑depth investigation by experienced adjusters, eliminating weeks or months of bureaucracy and paperwork for legitimate claims.
“If you look at the value chain of claims from the very first notification of loss all the way through settlement, a lot of judgment calls have to be made,” says David Hollander, global insurance leader at Ernst & Young. “AI can help ensure that the most effective claim handlers are handling the right claims.”
In chatbots we trust?
AI can also improve the customer experience. For decades, insurers have been plagued with low customer satisfaction rates. In 2015, just 1 in 3 customers (29%) reported being happy with their insurance provider, according to an Accenture survey of 23,000 consumers.
Thanks to machine learning tools applied at various points across the claims process, many insurers are now better equipped to mine insights into what their customers want and deliver an experience that matches.
When Chubb integrated machine learning into its claims processes, it began to analyze audio of all claims calls, giving managers a wealth of data about how service reps dealt with customers in high‑stress situations. “We went into AI to do fraud protection, and we came out with QA on customer interactions,” said Andrew Pelcin, Chubb’s vice president of claims data analytics, speaking at a recent recent InsuranceNexus webinar on claims automation.
Companies like Metromile are applying these insights to build chatbots that can handle calls with sensitivity in the aftermath of car accidents and other difficult, unfolding situations.
“When we think about claims handling, the priority is to focus on the customer,” said Amrish Singh, Metromile’s enterprise product head, during the webinar. “They’re at a vulnerable time in their life—they’ve suffered a loss. We ask between 50 and 100 questions in a claims call, but we ask them in such a way that the customer isn’t spending more than 5 to 10 minutes.”
Increasingly, chatbots such as Geico’s Kate and Allstate’s ABIE are fielding claims calls in place of human reps. In 2017, insurance companies invested $124 million in chatbot technology, more than any other industry, according to a recent Tata Consultancy Services study.
Some insurance providers have started automating virtually all claim processing. China‑based Zhong An Insurance launched in 2013 as an online‑only insurer and has since sold more than seven billion insurance products to some 420 million customers. Zhong An’s chatbots handle 97% of all customer interactions.
However, most insurance customers aren’t quite ready to deal exclusively with machines. More than half (55%) prefer a mix of AI and human interaction, according to a 2018 Capgemini survey of 10,000 consumers.
“Claims are a very emotional time,” Hollander says. “You can use a chatbot to allow people to have their questions answered quickly or to check the status of a claim. But if they want to speak to somebody, you have to get them to a human being.”
Crossing the digital divide
For the sector’s major players, AI adoption in claims management still poses challenges for a variety of reasons: legacy products, siloed data, or corporate cultures that are hard‑wired to resist change.
“Most people, you buy an iPhone and three years later you’re done with it,” Retzlaff says. “But if you buy an insurance product, companies are supporting it for 20, 30, 50 years. One company I worked with had policies that dated back to the 1800s.”
For many traditional insurers, the road to transformation requires a detour. Rather than blowing up their operating models, major companies like Allstate, MetLife, and MassMutual have launched incubator ventures to launch new products.
In 2017, Liberty Mutual started Solaria Labs to develop AI solutions, including a machine learning tool that can make accurate car‑repair estimates after being trained on thousands of photos of auto crashes.
Increasingly, the big insurers face competition from insurance tech startups like Lemonade (home), Tractable (auto) and Fabric (life). The new entrants are backed by major venture capital investment—$2.3 billion in 2017 alone.
Could one of these newbies emerge as the Uber of insurance? Maybe, says Hollander—but the big insurers won’t go down without a fight.
“They’re watching all of the insurance techs very carefully,” he says. “They get it. And in three to five years, you’ll see a bigger gap between winners and losers.”